The American Federation of Teachers and Microsoft have signed a legally binding safety and privacy standard. It protects classroom data from being used to train artificial intelligence. This contract-based agreement bypasses stalled federal regulations, which allows school districts to demand these protections directly from tech providers. But because the protections do not apply automatically, local school boards must actively negotiate them into individual contracts.
What Happened
On September 9, 2026, Microsoft became the first major AI developer to sign a binding memorandum of agreement with the National Academy for AI Instruction, a joint initiative between the American Federation of Teachers (AFT) and industry leaders. Microsoft agreed not to use student or teacher data to train its AI models. It also banned student tracking and automated AI decisions without human oversight. The agreement defines "Student Data" broadly, covering grades, names, user prompts, AI-generated outputs, behavioral patterns, and metadata. The AFT is negotiating similar terms with other companies, including Anthropic and OpenAI.
The Bigger Picture
This agreement addresses a major regulatory gap. Traditional privacy laws like the Family Educational Rights and Privacy Act (FERPA) were written decades ago. They fail to address the risks of machine learning, as previously reported. Instead of waiting for Congress, school districts can use contract law to protect students.
But relying on individual contract negotiations creates a divide. School districts spend billions of dollars on AI tools but struggle to evaluate their safety and value. Wealthy districts have specialized legal teams to write custom agreements. Smaller districts often lack the administrative capacity to demand and enforce these changes.
The technical premise of "anonymizing" data is also failing. Under the Microsoft agreement, the company can still collect "de-identified" usage data to improve its products. Computer science research shows that modern AI easily re-identifies individuals from supposedly anonymous datasets by cross-referencing public information, writing styles, and location details. A 2026 study published on arXiv found that autonomous AI agents re-identified over 70% of individuals in spatio-temporal datasets for pennies per target. When school devices track coordinates and writing styles, anonymization is no longer a reliable safety net.
What This Means for Families
This means your child’s essays, chat prompts, and digital behaviors are safer, but only if your local school board takes action. If your district uses Microsoft classroom tools and fails to amend its contract, your child's data remains subject to standard terms of service. Students in proactive, well-funded districts will enjoy strict AI privacy, while students in under-resourced districts will not.
What You Can Do
First, ask your local school board if they have requested to add the National Academy for AI Instruction’s privacy language to all Microsoft contracts. These contracts must be made available within 90 days of the agreement.
Second, urge state lawmakers to adopt model AI procurement guidelines. Steps taken in states like Idaho and Maryland show how states can keep local districts from having to negotiate complex technical terms individually.
Finally, review your district's software purchases. Use public databases or school board minutes to track vendor agreements, like those for SchoolAI. Make sure they contain strict provisions against using student inputs to train corporate AI models.